Thread by @asymmetricinfo

I'm afraid that I'm working on something else right now, so I really can't do this justice, but I'll briefly note that this is a case study in what Scott Alexander dubbed "isolated demands for rigor". slatestarcodex.com/2014/08/14/bew…

https://twitter.com/ryanlcooper/status/1169623031234355200

Mr. Cooper criticizes the various weaknesses with Dobkin et al: the fact that it is only in California, the fact that it only captures people who were hospitalized for a serious illness, the fact that it excludes the elderly and children.

These are all real potential weaknesses! They are noted by the authors. They were also noted by me, in my writeup: washingtonpost.com/blogs/post-par…

He is considerably less loquacious about the much larger methodological problems of Warren et al.

Such as: the relatively tiny sample (survey n of ca 1000 compared to observational n of half a million); response bias; a definition of medical bankruptcy that will sweep up any bankruptcy in which any non-trivial medical bills are involved, even if other debt dwarfs the md bills

He also doesn't seem to understand what Dobkin et al actually did.

This is just ... wrong. It's not a matter of opinion. It's a factually incorrect description of how the study worked.

This is what the study did:

They're establishing a relative risk for filing bankruptcy after a hospitalization. Mr. Cooper's scenario would be included in that relative risk.

Now as I say, there are obviously going to be some BKs they don't capture. They exclude children and the elderly--though, I quote the authors: "in other work we found that hospitalizations have no effect on bankruptcy rates among the elderly" (unsurprising; they're all insured.)

Also the elderly tend not to have garnishable wages, and their assets tend to be mostly judgment-proof home equity, pension and retirement accounts, which might make BK unnecessary even if bills caused a bunch of financial strain.

Point is, his "considered judgment" obviously wasn't considered quite hard enough.

Now, onto his other point: goalpost moving.

He looks at other studies which suggest results in the range of 25% of all BKs due to medical bills. I could quibble with methods and explain why I prefer Dobkin, but I won't, and here's why: I think, say, 20% of all BKs is well within the reasonable range of medbill-driven BKs.

I think Dobkin et al sets a floor; I don't think it's the ceiling. As I said in my own write-up, that number is certainly too low for all medical bks; the problem is, if you quadruple it, you're still not even in striking distance of Warren et al. Warren et al are just too high.

Or at least they're too high if the question we are trying to answer is "What percentage of bankruptcies could be prevented with a single payer health care plan?"

Since Sanders was quoting the figure in support of M4A, and Himmelstein/Wooldhandler are quoting it in support of their advocacy of a national health care program, presumably, that is the question that matters!

Mr. Cooper attempts to elide this by saying "The number's high, so who cares if it's 200k or 500k?"

I dunno, who cares if it's 200k or 1?

This is a singularly bizarre defense. Yes, eventually it's stupid to quibble over small differences. A factor of 2 is not a small difference

And all this started over a dispute with our fact-checker, who did not say that the number was zero, so proving that it is non-zero doesn't win the argument. He said 500k was too high. Mr. Cooper seems to agree it's too high. Well, why are we still arguing?

I mean, someone has been tweeting at me--repeatedly--that the correct number is "somewhere between" me and Bernie Sanders, citing Mr. Cooper. My only argument is that the number is not 500K. What is "between" 500k and not-500k. An infinitely infinitessimal repeating series?

I don't think Mr. Cooper has supported the Sanders position. Nor his own initial claim that I dismissed Sanders based only on my "extremely vague priors". I don't think he's even trying to support those statements any more, so as far as I'm concerned, our dispute is at an end.